arXiv:2604.20121v3 Announce Type: replace
Abstract: Range-filtered approximate nearest neighbor search (RFANNS) is increasingly critical for modern vector databases. However, existing solutions suffer from severe index inflation and construction overhead. Furthermore, they rely exclusively on CPUs for the heavy indexing and query processing, significantly restricting the throughput due to the limited memory bandwidth and parallelism. In this paper, we present Garfield, a GPU-accelerated framework for multi-attribute range filtered ANNS that overcomes these bottlenecks through designing a lightweight index structure and hardware-aware execution pipeline. Garfield introduces the GMG index, which partitions data into cells and builds local graph indexes. It guarantees linear storage and indexing overhead by adding a constant number of cross-cell edges. For queries, Garfield utilizes a cluster-guided ordering strategy that reorders query-relevant cells, enabling a highly efficient cell-by-cell traversal on the GPU that aggressively reuses candidates as entry points across cells. To handle datasets exceeding GPU memory, Garfield features a cell-oriented out-of-core pipeline. It dynamically schedules cells to minimize the number of active queries per batch and overlaps GPU computation with CPU-to-GPU index streaming. Extensive evaluations demonstrate that Garfield reduces index size by 4.4x, while delivering 119.8x higher throughput than state-of-the-art RFANNS methods.
Science Journals
arXiv:2607.08960v1 Announce Type: new
Abstract: Warehouse operations are governed by Standard Operating Procedures (SOPs) that encode complex, multi-system decision logic, which must be executed reliably under strict time constraints, yet LLM agents lack mechanisms to enforce procedural compliance and degrade under the context overload full SOP specifications introduce. We present Eluna, a production-deployed agentic system for reliable SOP execution. Eluna is a graph-guided, multi-agent framework that encodes SOPs as directed acyclic graphs with progressive disclosure and delegates independent tasks to parallel sub-agents, each with persistent code execution and live data access. To meet production latency and accuracy needs, we use asymmetric episodic distillation where a strong teacher is improved through episodic error memories, then a smaller student is fine-tuned on the corrected trajectories with memory stripped, internalizing corrections without inference-time overhead. On a 13-task benchmark and two production applications, our fine-tuned models match or exceed their teacher, beat all larger off-the-shelf baselines, and reach 94% expert agreement on the ticket processing application.
arXiv:2605.17676v2 Announce Type: replace
Abstract: Amid growing concern about information quality and credibility in digital media environments, researchers and educators still lack a concise, comprehensive yet psychometrically sound instrument for tracking the competencies that help people navigate this landscape. This article develops the Digital Media and Information Literacy Scale (DMILS), a robust and multidimensional measure that distinguishes domain (digital vs. information/news), competency type (knowledge vs. skill), and is measured through both subjective and objective items. Through two empirical studies with three nationally matched samples in the United States and Singapore (N = 1,498), we developed an 18-item self-report battery and 16-item objective knowledge questions, showing strong structural, convergent, and predictive validity, along with a short form (8 self-report and 8 objective items). By offering a parsimonious yet multidimensional yardstick, DMILS enables rigorous evaluation of media literacy interventions and supplies a common metric for cross-national research, critical for building an information ecosystem resilient to mis- and disinformation.
arXiv:2607.09155v1 Announce Type: new
Abstract: We introduce UISTful, a system that turns reading activity into a collective portrait of a scholarly community. Readers explore a semantic globe of UIST papers and authors while the system records private reading traces that can be reviewed, reflected upon, curated, and published for others to replay. Inspired by the information fl\^aneur, UISTful treats a reading trace as a camera through which readers frame and interpret what they read, casting reading as a creative and authorial process. Shared traces display the plurality of interpretations composed across the same scholarly landscape, while collective trace views reveal paths and concentrations of attention across the community, inviting UIST to see itself as an interactive system of papers, authors, readers, and their exchanges.
arXiv:2607.09163v1 Announce Type: new
Abstract: Gr\"obner basis computation incurs heavy computational overhead, especially under lexicographic order. F5 and its GVW variant dominate efficient field-based Gr\"obner basis solving. The proper basis algorithm offers a parameterized ideal computation framework without leveraging modern signature-based optimizations. This work presents the Proper-Cover algorithm for zero-dimensional polynomial ideals by combining GVW's cover optimization over signature with the proper basis theory. We generalize signature, cover, POT ordering, reduction and S-pair concepts to parameterized coefficients, design a two-phase algorithm with compatible factor construction and hungry refinement, and rigorously prove termination and output correctness. Accordingly, we propose a new framework for the efficient computation of polynomial ideal bases. Benchmark results show that Proper-Cover surpasses F5 under all monomial orderings and delivers clear speedups over GVW for lexicographic (plex) order.
arXiv:2607.07414v2 Announce Type: replace
Abstract: Address clustering is an important technique in blockchain forensics, widely employed by law enforcement to trace illicit crypto asset flows. The multi-input heuristic (MIH), which clusters addresses potentially associated with the same entity, is the most widely used. Yet, despite its broad adoption, the MIH has rarely been evaluated against reliable ground truth data. We implement a reusable evaluation framework covering nine established metrics and apply it to ground truth address-to-entity mappings obtained directly from European crypto asset service providers under legally mandated reporting obligations. When evaluation is restricted to reported addresses, the MIH appears strong at dataset level: we observe no mergers between reported services and recover same-service address pairs with recall 0.71. However, this result is driven by one large service and ignores unlabeled addresses absorbed into full clusters. Metrics that assess the full clusters show substantially lower precision and recall (0.36 and 0.44), meaning that services are often only partially recovered or embedded in larger clusters. Entity-level results further reveal near-complete failures for some services. When MIH-based clusters are used to support criminal suspicion, preliminary seizure of crypto assets to secure later forfeiture/ confiscation, or as evidence in trial proceedings, prosecutors and judges must account for the heuristic's metric-dependent and entity-dependent reliability.
arXiv:2607.08864v1 Announce Type: new
Abstract: This paper addresses the problem of observer design for a class of linear descriptor systems affected by a certain class of unknown unmatched disturbances. The objective is to estimate the components of the state vector, as well as the unknown parameters of the unmeasured disturbance. To solve this problem, structural assumptions are introduced under which an observer for the dynamic part of the state vector is constructed. Then, based on the obtained state estimate, the disturbance signal is reconstructed, and its unknown parameters are identified. A new parameterization method is proposed for a class of disturbance input signals that depend nonlinearly on unknown parameters, making it possible to obtain a linear regression in the corresponding unknowns. Numerical simulations are presented to demonstrate the effectiveness of the proposed procedures.
arXiv:2607.08896v1 Announce Type: new
Abstract: We describe our entry to the ICIP 2026 Grand Challenge on Extreme In-the-Wild License Plate Super-Resolution (XLPSR), which scored 9.73 wECR on the public validation leaderboard. The system pairs a Hybrid Attention Transformer super-resolution (HAT) front-end with an ensemble of two scene-text recognisers (PARSeq-S and CLIP4STR-B) and a confidence-weighted character-voting scheme that abstains on uncertain positions. We treat XLPSR as a recognition task gated by image legibility: the SR step exists to lift characters out of sub-pixel territory, and the asymmetric scoring rule (+2 / -1 / 0) is exploited explicitly through abstention. Our pipeline runs in 1.7 s per sequence on RTX 3090 (max 2.7 s, p99 2.4 s), well under the 60 s/sequence Docker budget.
arXiv:2607.09241v1 Announce Type: new
Abstract: Compact structures of intense tangential Reynolds stress (Q events) are well-known components of wall-bounded turbulence, and have been shown to be coherent because they approximately govern their own evolution. It has therefore often been assumed that they also are causally significant, in the sense that they explain the evolution of the incoherent component of the flow, which could thus be modeled as a superposition of structures. Since strong events typically only fill a small percentage of the total flow volume, this is also cited as a reason for considering structures as targets for efficient engineering flow control. This paper shows that the causality assumption does not hold in general. Only about half of the structures identified in the flow are causally more significant than an equivalent volume of incoherent turbulence. To explain this variability, feature-based analysis and conditional averaging are performed. For wall-attached Q2 events and wall-detached structures, causally enhanced events are characterized by elevated strain rate and spanwise vorticity. These signatures are traced to intense upstream strain regions generated by the interaction of Q4-like motions impinging on Q2 events. For wall-attached Q4 events, the dominant indicators are instead the wall-normal and streamwise vorticity components; enhanced causal significance is associated with strong wall-normal vorticity. These findings show that quadrant events cannot be treated as a dynamically homogeneous class in causal analyses. Their causal significance depends strongly on the local flow environment, emphasizing the need to interpret coherent structures in terms of their interactions with the surrounding turbulence.
arXiv:2607.08642v2 Announce Type: replace
Abstract: Speculative decoding accelerates LLM inference by drafting several tokens and verifying them in parallel. Block-diffusion drafters such as DFlash produce a draft block in one pass but model only per-position marginals, and best-first tree methods such as DDTree expand candidate trees from those marginals. The released Domino drafter adds a GRU-based causal correction that makes each draft token distribution path-dependent, a structure DDTree's factorized formulation cannot represent. We introduce DominoTree, a training-free best-first draft tree scored by Domino's conditional, non-factorized correction along each root-to-node path, made practical by restricting the per-node correction to a candidate top-M set. On Qwen3-4B across eight benchmarks, DominoTree reaches up to 6.6x speedup over autoregressive decoding and the highest mean accepted length of any evaluated method, up to 10.7 tokens per round, at every tested temperature. DominoTree constructs its tree with a GPU-native CUDA-graph builder that is bit-identical to a reference Python implementation, so acceptance is unchanged, while keeping per-round tree construction cheap. With this builder as default, DominoTree improves throughput over the released Domino decoder, the drafter it builds on, at every tested temperature: 9% to 10% overall on Qwen3-4B and up to 22% on Alpaca. It also outperforms DDTree and CaDDTree at every tested temperature, not only under greedy decoding. On Qwen3-8B, DominoTree keeps the highest accepted length at every temperature and gives a 24% throughput gain over DDTree at T=0; at higher temperature its edge over DDTree and CaDDTree narrows to a tie and a small loss, while its aggregate gains over DFlash and Domino persist.
arXiv:2604.01206v2 Announce Type: replace
Abstract: We present RELISH (REgression with a Latent Iterative State Head), a novel, lightweight architecture designed for text regression with large language models. Rather than decoding numeric targets as text or aggregating multiple generated outputs, RELISH predicts scalar values directly from frozen LLM representations by iteratively refining a learned latent state through cross-attention over token-level representations, and then mapping the final state to a point estimate with a linear regressor. Across six datasets, four LLM backbones, and two LLM training regimes, RELISH consistently outperforms prior baselines from all three major LLM regression families, including autoregressive decoding, regression-aware inference, and existing predictive head methods. Despite these gains, RELISH remains highly parameter-efficient, requiring only $\sim$3.4-3.7M trainable parameters across frozen LLM backbones (only 0.01-0.04$\%$ additional overhead), far less than LoRA-based alternatives that grow with model size (0.26-0.42$\%$). Our code is available at https://github.com/SamSoup/RELISH.
arXiv:2605.18581v2 Announce Type: replace
Abstract: The low-field carrier mobilities in <100> silicon were quantified as a function of the 1$\,$MeV neutron-equivalent fluence up to $10^{18}\,$cm$^{-2}$ and for temperatures between 230$\,$K and 260$\,$K. Current measurements were fitted using a mobility model for scattering at ionized impurities. Technology-aided design (TCAD) simulations were compared to measurements and used to estimate the carrier concentrations, which are parameters in the fit. The fit model describes the data very well, both as a function of fluence and the temperature. At a fluence of $6 \cdot 10^{17}\,$cm$^{-2}$, which is expected for the innermost detector layers at the proposed Future Circular Hadron Collider (FCC-hh), the sum of the mobilities of electrons and holes was found to decrease by $\sim60$%.
arXiv:2607.08963v1 Announce Type: new
Abstract: Continual counting under pure differential privacy is one of the simplest and most well-studied problems in the continual observation model. Nevertheless, an asymptotic gap remains between the best known upper and lower bounds for maximum squared error and mean squared error: the upper bound is $O(\epsilon^{-2}\log^3 n)$, while the lower bound is $\Omega(\epsilon^{-2}\log^2 n)$, for both error metrics. The best known constant in the upper bound is achieved by the $k$-ary tree mechanism with the subtraction trick, due to Andersson, Pagh, Steiner, and Torkamani (FORC 2025). In this work, we improve the leading constant in the maximum squared error and the mean squared error. Our approach uses a general matrix factorization mechanism, yielding an improved bound for pure-DP continual counting that does not rely on a tree-based construction. The mechanism starts from a good-quality low-dimensional factorization, obtained via gradient-based optimization, and gives an explicit matrix construction that lifts this factorization to arbitrarily large dimensions, further improving its error guarantees. We offer an efficient algorithmic implementation of our mechanism. On the lower-bound side, we prove an $\Omega(\epsilon^{-2}\log^3 n)$ lower bound for the class of factorizations whose matrices have entries in $\{0,1\}$, matching the upper-bound asymptotics for this class. This class includes the binary tree mechanism and $k$-ary tree mechanisms without the subtraction trick. Extending this lower bound to arbitrary matrix factorizations, and beyond the matrix mechanism altogether, remains an open problem.
arXiv:2607.09327v1 Announce Type: new
Abstract: We study the problem of computing an envy-free (EF) allocation of $m$ indivisible items among $n$ agents when items come in three distinct types. Each agent holds additive valuations over item types that may be positive (goods), negative (chores), or mixed. We present the first subexponential-time algorithm with running time time $(n \cdot m)^{O(\sqrt{n})}$ that finds an EF allocation whenever one exists, or correctly reports that none exists. Our approach exploits a geometric representation of EF allocations as convex polyhedra in $\mathbb{R}^3$ and applies Miller's planar cycle-separator theorem to recursively decompose the agent set into balanced subgroups. We further extend the algorithm to handle agents whose allocations are fixed in advance, preserving envy-freeness across all agents.
arXiv:2605.22220v2 Announce Type: replace
Abstract: Orbital Angular Momentum (OAM) of light is a promising degree of freedom for next-generation communication. By exploiting the orthogonality of OAM modes, multi-channel division enables a linear increase in communication performance proportional to the number of available modes. However, the multiplexing and demultiplexing of each superposition state remain essential yet complex processes. Demultiplexing has been established through spatial-domain methods that require additional optical elements such as gratings and apertures, which can decrease communication efficiency and accuracy under various conditions. In this paper, we propose a demultiplexing method under a single intensity profile by orthogonality restoration of OAM pair states. This method can be applied directly to an OAM multichannel communication system without additional receiver-side optical structure. We present simulation results of our method under various conditions.
arXiv:2605.05882v2 Announce Type: replace-cross
Abstract: Artificial-intelligence systems are becoming ubiquitous in society, yet their predictions typically inherit biases with respect to protected attributes such as race, gender, or age. Classical fairness notions, most notably Statistical Parity (SP), demand that predictions be independent of the protected attributes, but are overly restrictive when these attributes influence mediating variables that are considered business necessities. Recent causal formulations relax SP by distinguishing allowed from not-allowed causal paths and by complementing SP with Predictive Parity (PP), requiring the predictor to replicate the legitimate influence of business-necessities. Existing path-based definitions are mainly practical when applied to categorical attributes. This paper introduces a new framework for fairness in structural causal models that is tailored to continuous protected attributes. We formalize SP and PP through path-specific partial derivatives, establish conditions under which these criteria coincide with prior causal definitions, and characterize when a fair predictor, one that satisfies SP along not-allowed paths while achieving PP along allowed paths, exists. Building on this theory, we propose a fair tuning algorithm that either constructs such a predictor or, when not possible, allows for a trade-off between SP and PP. We present experiments on simulated and real data to evaluate our proposal, compare it with previously proposed methods, and show that it performs better when PP is considered.
arXiv:2607.09469v1 Announce Type: new
Abstract: We introduce a new notion of distance between two graph states $|G\rangle$ and $|G'\rangle$ on the same set of qubits. This distance is the minimum number of ancilla qubits in a graph state $|\widehat{G}\rangle$ from which both $|G\rangle$ and $|G'\rangle$ can be ``easily prepared''. (When preparing graph states, we are only allowed to use one-qubit Clifford gates, one-qubit Pauli measurements, and classical communication.) We give a graphical description of this distance through the lens of vertex-minors. We then show how this distance yields quantum network analogs of many graph edit-distance problems.
Using this framework, we develop classical algorithms for identifying the ``highly entangled clusters'' of a graph state $|G\rangle$. The ancilla integrity problem asks, given a graph $G$ and integer $k$, for the minimum -- over all graph states $|G'\rangle$ with distance at most $k$ from $|G\rangle$ -- of the maximum component size of $G'$. Up to a factor of $2$ in the number of ancilla qubits, this problem is equivalent to rank integrity, where the distance between $G$ and $G'$ is instead the minimum rank of the sum of their adjacency matrices over $\text{GF}(2)$. We prove that rank integrity is XP parameterized by $k$. We also prove the complementary hardness result that rank integrity is W[1]-hard in $k$. Finally, we give an explicit $\mathcal{O}(n^6)$-time algorithm for ancilla integrity when $G$ has $n$ vertices and $k=1$.
arXiv:2607.08954v1 Announce Type: cross
Abstract: We study nonasymptotic convergence of primal-dual methods for a class of nonconvex constrained optimization problems with a convex-composite structure. In this class, both the objective and the functional inequality constraints are given by convex Lipschitz outer functions composed with smooth nonlinear inner mappings. The analysis is complicated by constraint violation in a nonconvex functional inequality system and by the lack of an a priori bound on the multipliers. To address these issues, we restrict the dual variable to an auxiliary compact set and analyze a smoothed prox-linear augmented Lagrangian method through a nonsmooth nonconvex-concave minimax reformulation. The main contribution is a finite-time mechanism for converting stationarity of the truncated minimax problem into a KKT certificate for the original constrained problem. We show that, for a sufficiently large penalty parameter, all but a controlled number of iterates enter a near-feasible region. On this region, a local conic regularity condition uniformly bounds the associated prox-linear multipliers and thereby makes the artificial dual truncation inactive at the selected iterates. Building on this mechanism, we establish explicit convergence rates for the proposed method in terms of the KKT residual. With dual regularization, a global dual error bound together with a bias-balancing argument gives an $O(K^{-1/3})$ rate. In the unregularized case, under additional local structural assumptions including piecewise linearity of the outer functions, a local dual error bound yields the sharper $O(K^{-1/2})$ rate.
arXiv:2607.07740v2 Announce Type: replace
Abstract: Modern LLMs are increasingly deployed in long-context applications such as retrieval-augmented generation, repository-level coding, and agentic workflows whose accumulated reasoning and tool traces routinely push the input an order of magnitude past the pretraining window, making zero-shot context extension the dominant deployment path for open-weight checkpoints. The dominant zero-shot methods (YaRN, Self-Extend, DCA) fix a single rescaling factor up front, so an aggressive factor sacrifices short-context fidelity while a conservative one breaks down at long contexts; recent length-aware variants adapt the mapping, but with a fitted or distance-dependent schedule. We propose Jet-Long, a tuning-free zero-shot method that pairs a local RoPE-faithful window with a long-range window whose rescaling factor adapts dynamically to the current sequence length via a parameter-free analytic schedule, recovering the base model exactly at short inputs while extrapolating cleanly at long ones. An inclusion-exclusion attention merge and an on-the-fly RoPE correction rotation make the bifocal construction essentially free at inference; fused into a single CuTe kernel, long-context prefill reaches up to $1.39\times$ FA2 throughput on H100 (approaching the Hopper-only FA4), and single-batch generation incurs $\le 4\%$ overhead at every length. On Qwen3-1.7B/4B/8B up to 128K context, Jet-Long leads RULER by $+4.79$/$+2.18$/$+2.03$ pp over the strongest baseline at 1.7B/4B/8B, achieves the best overall accuracy on HELMET-RAG (a benchmark identified by HELMET as the most efficient predictor of downstream long-context performance) and attains the lowest PG-19 perplexity. Jet-Long also generalizes to hybrid attention architectures such as Jet-Nemotron for further long-context improvement without retraining, and remains hyperparameter-resilient for ease of deployment.
arXiv:2607.09149v1 Announce Type: new
Abstract: The rapid growth of scientific publications makes scholarly taxonomies quickly obsolete. We study taxonomy maintenance in the wild, a new problem that moves beyond static construction by continuously adapting taxonomies to evolving scholarly repositories, such as arXiv, for a given research topic. We propose GIST, a robust framework for maintaining evolving taxonomies. Unlike purely LLM-centric approaches, GIST grounds structure induction in expert-curated evidence by extracting partial hierarchies from the "Related Work" sections of papers. It integrates these partial taxonomies into a unified global taxonomy in a geometric box-embedding space, where box containment encodes the inductive bias of is-a relations. To connect semantics with geometric structure, GIST learns a bidirectional mapping between word embeddings and box embeddings. For efficient incremental updates, GIST uses novelty-aware coreset selection to update the model with representative historical signals and new evidence, avoiding costly full retraining. To handle high-velocity paper streams under user-specific token budgets, GIST further combines a hypothesized concept generator with a cost-effective evidence retrieval module. Experiments on real-world arXiv datasets show that GIST outperforms state-of-the-art baselines, improving Node F1 and Edge F1 by 11.0% and 13.1% over the strongest baseline while requiring only 9.6% of its runtime and 12.7% of its monetary cost.
arXiv:2606.00726v2 Announce Type: replace
Abstract: Strong reasoning depends not only on model knowledge but also on how effectively cognitive behaviors are deployed during generation. Existing methods often rely on explicit behavior-level control, making them insufficiently adaptive when failures and required corrections vary across reasoning states, tasks, and models. To this end, we propose Latent Reward Steering (LRS), an adaptive inference-time framework that promotes cognitive behaviors by optimizing the sparse-autoencoder (SAE) latent states that implicitly carry them. Rather than relying on predefined cognitive behaviors or steering directions derived from them, LRS trains a latent reward model on reasoning traces by final answer correctness to estimate the quality of intermediate latent states. During inference, reward gradients provide state-specific correction directions for fragile latent states, while a reward and confidence gate restricts intervention to states the reward signal flags as fragile. Experiments on multiple reasoning LLM backbones and benchmarks show that \ours consistently improves performance over various baselines, and post-hoc analyses further indicate that \ours implicitly promotes good cognitive behaviors that fix the original reasoning errors. Code is available at: https://github.com/jiakanglee/Latent-Reward-Steering.
arXiv:2601.21764v3 Announce Type: replace
Abstract: We introduce a method for solving Hamilton--Jacobi equations, both inviscid and viscous, by minimizing the squared residuals of monotone finite-difference discretizations on grids of varying resolution. The method is designed to leverage neural networks and modern GPUs to solve these equations in higher dimensions; consequently, the setting for our analysis is the minimization of the residual functionals via gradient-based optimization. We establish a well-posedness theory for this approach: any critical point of the finite-difference loss solves the monotone scheme together with the prescribed Dirichlet boundary conditions, and the error of an approximation is controlled by its residual. We then derive the rate of convergence of the gradient flow that minimizes the residual in several settings, noting that the rate may depend on the grid resolution and the domain dimension. Building on this foundation, we propose a multi-level training algorithm that exploits the faster convergence available on the coarser grids of the discretization. Combined with the convergence theorem of Barles and Souganidis for monotone and consistent schemes, our results guarantee convergence to the unique viscosity solution as the grid is refined. We illustrate the approach on eikonal equations, level-set problems, and a Hamilton--Jacobi--Isaacs equation arising from a stochastic differential game, in dimensions up to eight.
arXiv:2607.09526v1 Announce Type: new
Abstract: Foundation models are reshaping computational pathology, yet their capabilities remain shaped by pretraining objectives, data sources, and spatial scales, fragmenting complementary expertise across separate backbones. Here we present ALICE, a unified foundation model trained through multi-stage agglomerative distillation that sequentially distills eight vision-only, vision-language, and slide-level teacher models into dedicated modules of a single backbone. ALICE is pretrained on 24,985,184 tile-level pathology images and 155,604 high-resolution images, and evaluated across 21 task scenarios, 96 downstream tasks, and 48 data sources, spanning region-of-interest tissue analysis, vision-language multimodal evaluation, and whole-slide clinical assessment. In all three evaluation settings, ALICE achieved the best average rank among task-matched pathology foundation models. These results demonstrate that agglomerative distillation can consolidate complementary capabilities from specialized models into a unified backbone for broad computational pathology applications. The model is available at https://github.com/WonderLandxD/ALICE.
arXiv:2602.04078v2 Announce Type: replace
Abstract: Deep learning has achieved remarkable success across a wide range of domains, significantly expanding the frontiers of what is achievable in artificial intelligence. Yet, despite these advances, critical challenges remain -- most notably, ensuring robustness to small input perturbations and generalization to out-of-distribution data. These critical challenges underscore the need to understand the underlying fundamental principles that govern robustness and generalization. Among the theoretical tools available, Lipschitz continuity plays a pivotal role in governing the fundamental properties of neural networks related to robustness and generalization. It quantifies the worst-case sensitivity of network's outputs to small input perturbations. While its importance is widely acknowledged, prior research has predominantly focused on empirical regularization approaches based on Lipschitz constraints, leaving the underlying principles less explored. This thesis seeks to advance a principled understanding of the principles of Lipschitz continuity in neural networks within the paradigm of machine learning, examined from two complementary perspectives: an internal perspective -- focusing on the temporal evolution of Lipschitz continuity in neural networks during training (i.e., training dynamics); and an external perspective -- investigating how Lipschitz continuity modulates the behavior of neural networks with respect to features in the input data, particularly its role in governing frequency signal propagation (i.e., modulation of frequency signal propagation).
arXiv:2607.09183v1 Announce Type: new
Abstract: The groundbreaking development of generative artificial intelligence (AI) is rapidly boosting the ability to generate content such as images and videos, reshaping communication paradigms. This article introduces generative communications (GenCom), a novel paradigm for 6G networks in which large AI models (LAMs) drive semantic understanding, reasoning, and content generation, embedding these into the communication process. Unlike traditional systems that strictly pursue accurate bit transmission, GenCom enables transmitters to convey only minimal yet sufficient information, while receivers leverage shared generative priors and knowledge bases to synthesize the intended output. Communication is thus redefined as controlled generation rather than data reproduction. We formalize the concept of GenCom, clarify its AI-native and generation-driven properties, and present its core mechanisms. A two-layer GenCom architecture supported by key enabling technologies is proposed, and analysis of four representative application scenarios demonstrates that GenCom offers ultra-efficient transmission, semantic-level robustness, and new network functions. Finally, we outline future research directions, including foundational theory and real-time processing, highlighting a promising pathway toward 6G networks.